----- Forwarded message from GREG CAMPBELL
<[email protected]> -----
Date: Wed, 5 Mar 2014
07:36:12 -0500
From: GREG CAMPBELL
<[email protected]>
Reply-To: GREG CAMPBELL
<[email protected]>
Subject: Re: Shape analysis
without removing size as a factor?
To: "[email protected]"
<[email protected]>
Dear CT: Procrustes size does not remove
size as a factor. It is just another measure of size, it is not THE SIZE at
whose alter all must worship. It is a very strange measure of size for most
real-world applications, since it does not converge: using more landmarks does
not increase the precision of the size measure, but inflates it by the square
root of the number of landmarks. What effect serious differences in variance
between landmarks has on Procrustes size has not been examined as far as I know
(do sing out on this forum if you know it has), but (since variation along the
shaft will be much greater than variation around the epiphyses) I bet this could
also be a problem for you.
The measure of size should
be chosen by you,
depending on what you want to know about the tibia (compressive load, torsional
resistance, shear strength, mineral reserve capacity, metabolic demand in the
animal). Since growth is allometric (any of these loads or demands alters as the
size of the animal increases) size varies with shape (size and shape are the
two sides of the same coin). Since most allometry is what Stephen Jay Gould
called 'simple' (exponential, linear in log-scale), a good start is some
regression of the log-transformed data against some general measure of
size.
You may have already found what size to use:
you can examine your PCA or principal warps at various Procrustes sizes to see
what it is about that tibia that varies between sizes, and measure that bit of
the bone which varies. "Geometric landmark methods may be limited to defining
the morphological location and orientation of principal deformations to guide
the
choice of variables for more traditional analyses of linear distances" (Cadrin
2000, Reviews in Fish Biology and Fisheries 10: p 99). Frankly, that should
be enough for a good Masters paper.
Greg
Campbell
The Naïve Chemist
From:
"[email protected]"
<[email protected]>
To: [email protected]
Sent: Wednesday, 5 March 2014, 6:05
Subject: Shape analysis without removing size as a factor?
To: [email protected]
Sent: Wednesday, 5 March 2014, 6:05
Subject: Shape analysis without removing size as a factor?
----- Forwarded message from [email protected] -----
Date: Fri, 31 Jan 2014 20:33:52 -0800
From: [email protected]
Reply-To: [email protected]
Subject: Shape analysis without removing size as a factor?
To: [email protected]
----- Forwarded message from Celena Toon <[email protected]> -----
Date: Mon, 20 Jan 2014 15:36:52 -0500
From: Celena Toon <[email protected]>
Reply-To: Celena Toon <[email protected]>
Subject: Shape analysis without removing size as a factor?
To: [email protected]
Hello,
I've been working on my master's thesis that uses a geometric
morphometric approach to analyzing the human tibia and the _expression_
of sexual dimorphism. I've previously consulted this forum about
formatting my text files and it has been a wonderful help! After
conducting my analyses, I did not get the results expected and my
advisor wants me to seek other ways I could potentially analyze my
data to cover all my bases and make sure I'm not doing something
wrong. Using MorphoJ, I conducted a Procrustes fit, a principal
components analysis, and a discriminant function analysis. I know
that the Procrustes fit removes size as a factor, but is there a way I
could analyze my data in terms of both size and shape? Or should I be
approaching this differently?
Thank you,
CT
----- End forwarded message -----
----- End forwarded message -----
-----
End forwarded message -----
